In cellular wireless systems, Wideband Code-Division Multiple-Access (W-CDMA, or the European equivalent UMTS (Universal Mobile Telecommunications System)) is the radio-access technology available worldwide providing high-speed wireless data packet service, i.e. wireless internet access and multimedia service, as well as conventional voice services.
To enable more-efficient downlink data packet transmission, a closed-loop based rate control called High-Speed Downlink Packet Access (HSDPA) was adopted as the next UMTS release. HSDPA is based on a closed-loop rate-control scheme in which the downlink transmission data rate is changed as rapidly as channel variation by controlling rate-matching and modulation parameters.
Similar to HSDPA, a closed-loop-based rate control can be used for uplink data transmission so that UE can change the uplink transmission rate to compensate for a short-term channel condition. As illustrated in FIG. 1, uplink transmission rate of an individual user's UE is controlled by a base station (node B) by exchanging feedback signals in both uplink (rate request) and downlink (rate control).
User equipment (UE) may request a higher data rate to node B if the current maximum data transmission rate Rue (UE threshold) cannot meet a required Quality of Service (QoS) of data packets in its internal buffer. Otherwise, the UE can request to keep or reduce Rue in order to allow other UEs in a same cell to transmit at a higher data rate.
Having requests from all UEs in the cell, node B controls the data transmission rate Rnb of each individual UE based on additional internal information such as current uplink noise level. Initially, Rnb′ and Rue′, the respective starting transmission rates of node B and the UE (mobile phone in the cell of node B), are set to be equal to each other. Subsequently, node B decides on a new transmission rate Rnb, i.e. (Rnb′+dnb) to be used by the mobile phone, where dnb is a differentially-encoded control packet for changing the transmission rate of the mobile phone, and node B transmits dnb to the mobile phone. Assuming that the differentially-encoded control packet that the mobile phone detects in designated due, then due is equal to dnb only if the transmission from node B to the UE has proceeded without error; if an error occurs during the transmission, then due will not equal dnb.
There are two sources of feedback errors: (1) an uplink error in UE rate request, and (2) a downlink feedback error in the rate control command of node B. Both types of errors are inevitable because of factors such as limitations in transmission power, severe fading in wireless channels, etc.
Due to the fact that only node B has the authority to control the maximum data rate of the mobile phone, the uplink error is less of a problem than the downlink error. An uplink error incurs a slow adaptation of uplink rate control, while the latter breaks down the synchronization between Rnb and Rue. Loss of such synchronization causes the mobile phone, i.e. the UE, to transmit the data packet with excessively-high or excessively-low transmission power. With the former, there is a significant loss of manageability of uplink noise, while with the latter, the UE suffers low data throughput. These trade-offs in data transmission rates for W-CDMA are illustrated in FIG. 2.
FIGS. 3 and 4 further illustrate the foregoing comments. FIG. 3 is a general schematic illustration of the closed-loop communication between node B and a UE, i.e. a base station and a mobile phone, respectively. The UE cannot transmit at a higher data rate than the maximum rate which is set by node B when the radio link between the two is established. If the UE requires for transmission a higher data rate than the established maximum data rate, the UE can request the higher data rate by sending a request in an uplink, i.e. the lower part of FIG. 3. If the UE does not require the higher data rate, for instance due to small packet queue size, it can send a request to node B to reduce the maximum transmission rate. Node B manages multiple UEs to share the same radio resources efficiently by controlling the maximum data rates for all users in node B's cell. Node B can accept or reject any UE request for data transmission rate increase, dependent on overall cell level conditions.
FIG. 4 illustrates a general timing diagram of a closed-loop-based rate control scheme between node B and a UE. In frame ‘k’, node B recognizes that the rate request on the uplink feedback channel from UE of which the current maximum rate is Rkue, and after elapse of node B processing time Tbp, node B transmits an increased maximum data transmission rate (denoted by ‘U’ for UP) to the UE on the downlink feedback channel. After elapse of UE processing time Tbp, the UE increases its maximum data transmission rate to the new higher value allowed by node B. Then the process repeats in frame k+1, frame k+2, etc. In order to reduce the rate update delay (e.g. Td<10 ms), the duration of the UE request and the node B control command are assumed to be small (e.g. 2 ms) and to be located near the beginning or end of the feedback channel frame (e.g. slots 1 to 3 or slots 12 to 14, respectively). In FIG. 4, the UE has received the U (UP), D (DOWN) and K (KEEP, i.e. stay the same) data-transmission-rate commands without any error in downlink transmission.
FIG. 5 illustrates an example of an error occurring in the closed-loop-based transmission. Such error results in a “random walk.” The third rate command from node B, i.e. K (keep maximum data transmission rate the same), has been received at the UE as U (increase the maximum data transmission rate). All subsequent frames of communication between node B and the UE will have a mutual offset in their data transmission rates, i.e. random walking starts, even if no further error occurs between what node B transmits and what the UE considers it to be. Assuming even a low 1% error rate, the data rate control scheme breaks down after a few seconds of radio link connections due to random walking of the maximum data transmission rate between node B and the UE. An uplink (UL) error (from the UE to node B), may also result, but unlike the downlink (DL) error (from node B to the UE), the uplink error will only cause a delay and will not cause an error in the uplink data rate.
A number of matters must be considered in approaching a solution to the foregoing problem. Firstly, a solution that only reduces the probability of random walking is not suitable in the case of lengthy radio communications. Secondly, any solution should be radio wave spectrum efficient in order to guarantee more radio resources are allocated to data transmission per se rather than control of data transmission rate. Thirdly, a solution should avoid introducing a new physical channel which will impact on backward compatibility and design of new hardware, among other factors. Fourthly, it is desirable that there be no, or only minor, impact on the system's radio network controller (RNC), which controls essential communication between node B and the UE, since any involvement by the RNC necessitates additional expensive signalling between the RNC and node B, and also possibly between the RNC and the UE.
One conventional solution to the foregoing difficulty has involved explicit signalling of the maximum UE rate from node B to the UE. In this solution, node B explicitly transmits a new maximum data transmission rate Rnb per se with each data packet rather than only transmitting a differentially-decoded dnb bit stream representing a change in the rate with each data packet. This solution, which is illustrated in FIG. 6, eliminates random walking. However, the spectral efficiency of this solution is lower than for the differentially-encoded bit stream due to the larger number of signalling bits per rate-control command. Assuming that the whole range of values of the maximum data transmission rate may assume any of 32 levels, 5 bits are required per rate-control command sent from node B to the UE. Since the 5 information bits equate to a higher number of encoded bits (e.g. 20 encoded bits), sending those 20 bits with each rate-control command will necessitate that a new physical channel be added if the control command is to be kept below 2 ms in length. The rate-control round-trip delay will increase if the encoded rate control bits are spread over a frame of, say, 10 ms. A major difficulty with this solution is that the overhead is much higher than that involved with transmitting only a differential value.